From Cis-Regulatory Elements to Complex RNPs and Back

Abstract

Messenger RNAs (mRNAs), the templates for translation, have evolved to harbor abundant cis-acting sequences that affect their posttranscriptional fates. These elements are frequently located in the untranslated regions and serve as binding sites for trans-acting factors, RNA-binding proteins, and/or small non-coding RNAs. This article provides a systematic synopsis of cis-acting elements, trans-acting factors, and the mechanisms by which they affect translation. It also highlights recent technical advances that have ushered in the era of transcriptome-wide studies of the ribonucleoprotein complexes formed by mRNAs and their trans-acting factors.

mRNAs harbor abundant cis-acting elements that direct the assembly of ribonucleoprotein particles (mRNPs), which influence their posttranscriptional fate.

Translational regulatory mechanisms are based on two key principles: signal-dependent covalent modifications of general translation (initiation) factors and trans-acting RNA-binding factors (RNA-binding proteins [RBPs] and miRNAs) to alter the translational fate of mRNAs. The first cis-regulatory elements to be found in eukaryotic mRNAs were the upstream open reading frames (uORFs) of the yeast GCN4 mRNA (Mueller and Hinnebusch 1986) and the iron-responsive elements of mammalian ferritin mRNAs (Hentze et al. 1987; Leibold and Munro 1987). Coincidentally, they provided one example of each: cis-acting elements that function in the context of modified initiation factor activity (GCN4), or that serve as binding sites for the first translational regulatory proteins, the iron regulatory proteins (IRPs) (Hinnebusch 2005; Hentze et al. 2010), respectively. More than two decades later, translational control is recognized as a major control point for the flux from genetic information to shaping proteomes, and is even reported to be the predominant mechanism for the control of gene expression (Schwanhausser et al. 2011).

Initially not anticipated, mRNAs now have to be seen as linear yet structured arrays of numerous cis-acting elements, mostly in the 5′ and 3′ untranslated regions (UTRs) but probably spreading across the whole message. This situation is mirrored with regard to mRNA engagement with trans-acting factors, and mRNAs must therefore be examined as messenger ribonucleoprotein particles (mRNPs).

Cis-acting elements come in different flavors (Fig. 1). Hairpin or higher-order (e.g., pseudoknot) intramolecular mRNA structures can influence translation in their own right (i.e., without binding factors). It has been well documented that they can affect initiation, particularly when positioned in the 5′ UTR close to the cap structure (Pelletier and Sonenberg 1985; Kozak 1986), and elongation, being involved in numerous frameshifting events (Namy et al. 2004). Internal ribosome entry sequences (IRES) represent another important cis-acting element that typically occurs in 5′ UTRs but has also been reported to occur within the coding region of mRNAs (Holcik et al. 2000). In cellular mRNAs, IRES coexist with the 5′-cap structure and endow mRNAs with the potential to be translated under conditions in which cap-dependent translation is compromised (e.g., different forms of cell stress, apoptosis). A third category of common cis-acting elements comprises uORFs. They occur singularly or multiply within the 5′ UTRs of numerous mRNAs and influence the translation of the downstream major ORF, usually negatively. This effect can be exerted by the element itself, but appears to be used also for RBP-mediated translational regulation (see below). A notable exception is the GCN4 mRNA in yeast and ATF4 mRNA in mammals, where uORFs serve to promote the translation of the downstream major ORF under conditions of increased eIF2 phosphorylation (Hinnebusch 2005). Binding sites for regulatory RBPs or miRNAs can be combined on given mRNAs to yield translational outcomes that integrate multiple signals via the respective trans-acting factors. In a complementary concept, nearly all trans-acting factors bind to a multitude of mRNAs that frequently encode functionally related proteins, subjecting these families of mRNAs to coordinated, operon-like regulation (Keene 2007). Finally, RNA editing or modification (e.g., methylation) can provide an additional layer of regulatory intervention for cis-acting elements (Li et al. 2009; Squires et al. 2012; Zhang et al. 2012).

Figure 1.

Cis-acting elements that influence mRNA translation. The nearly ubiquitous 5′ ^m7GpppN cap structure (black circle) and 3′ poly(A) tail ((A)_n) strongly stimulate translation. Secondary structures (e.g., hairpin) and upstream open reading frames (uORFs) in the 5′ UTR usually inhibit translation. Internal ribosome entry sequences (IRES) stimulate translation independently of the cap structure. Binding sites for regulatory RNA-binding proteins or microRNAs (ovals) can provide positive or (more frequently) negative regulation. (Adapted and modified from Gebauer and Hentze 2004.)

The arrival of technologies to examine mRNAs and RBPs in a highly parallel, transcriptome-wide fashion places translational control at the heart of modern systems biology. Below, we discuss how system-wide studies and reductionist mechanistic experimentation must converge and complement each other for a deeper understanding of translational control.

THE mRNP AS A TEMPLATE FOR TRANSLATION AND TRANSLATIONAL CONTROL

Eukaryotic mRNAs pass through all stages of their life cycle from transcription, to processing, transport (including specific subcellular mRNA localization), translation, and decay, as assemblies of dynamic mRNPs rather than as “naked” nucleic acids (Martin and Ephrussi 2009; Moore and Proudfoot 2009; Sonenberg and Hinnebusch 2009). The implications of this notion are particularly profound for the interaction of the mRNA with the translation apparatus (translation factors and ribosome) as a template for protein synthesis. Parallels have previously been drawn between the processes of eukaryotic transcription and translation, likening the near-ubiquitous mRNA 5′-cap structure and 3′-poly(A) tail to constitutive promoter elements, whereas cis-regulatory motifs in the UTRs were viewed as mRNA-specific elements that act to control translation in a combinatorial and context-specific manner akin to transcriptional regulatory elements such as silencers and enhancers (Sachs and Buratowski 1997; Gebauer and Hentze 2004). This useful concept may be extended by an analogy between DNA in chromatin as the “real” template for transcription and the mRNA interacting with the translation machinery packaged into complex mRNP particles, which may even adopt additional states of “compaction,” for instance, in the form of repressive cytoplasmic foci such as stress granules and processing bodies (Anderson and Kedersha 2009; Balagopal and Parker 2009). Adoption of such an mRNP-centric view could provide similar conceptual breakthroughs for translation research as the chromatin model did for the transcription field.

At present, we are only scratching the surface in terms of understanding the composition of mRNPs, and the analogy with transcription may break down at the point of histones and nucleosomes. Even the most common RBPs (e.g., the hnRNPs) do not appear to organize mRNPs in a regular fashion as histones do with the chromatin template. There is a pressing need for identifying all cellular proteins interacting with mRNA within native mRNPs (“mRNA interactomes”), and then to study their organization and dynamic interplay. At least the determination of the first mRNA interactomes has now become feasible (see below).

NEW TECHNOLOGIES TO STUDY CIS-ACTING ELEMENTS AND TRANS-ACTING FACTORS

The complete description of an mRNA interactome in a given cellular condition requires cataloging the proteins that bind to each expressed mRNA and a high-resolution mapping of all respective binding site(s) on the mRNA. Conventional affinity copurification and low-throughput identification of binding partners have long been useful tools for characterizing individual examples and have started the journey along this road. However, such small-scale methods cannot cope with the magnitude of the new tasks at hand. Valiant attempts have been made to draw the power of bioinformatics into the challenge and to predict the occurrence of known cis-acting motifs across larger spectra of mRNAs. However, these approaches are limited by the degeneracy of the primary RNA sequence motifs involved (e.g., miRNA target sites) (Saito and Saetrom 2010) and/or by the important role that RNA secondary and tertiary structure plays in defining functional regulatory motifs (Parker et al. 2011; Zhao et al. 2011). The computational prediction of RBPs has helped significantly in the identification of new candidates, but it is by definition limited to proteins bearing known RNA-binding domains.

Identification of Cis-Acting Elements by Cross-Linking

A long existing but recently refined and popularized approach to study RBPs as well as the mRNAs and sites they bind to is to covalently cross-link the two to each other in vivo. Covalent cross-links can be induced chemically (Valasek et al. 2007) or by ultraviolet (UV) light. UV light of 254 nm cross-links the naturally photoreactive nucleotide bases, especially pyrimidines, and specific amino acids (Phe, Trp, Tyr, Cys, and Lys) (Hockensmith et al. 1986; Brimacombe et al. 1988). UV-cross-linking requires direct contact (zero distance) between protein and RNA and does not promote protein–protein cross-linking (Greenberg 1979). A recent version of in vivo UV cross-linking is called PAR-CLIP (photoactivatable-ribonucleoside-enhanced cross-linking and immunoprecipitation) (Fig. 2A) (Hafner et al. 2010). The photoactivatable nucleotide 4-thiouridine (4SU) is taken up by cultured cells and incorporated into nascent RNAs, and efficient cross-linking is induced by 365-nm UV light irradiation, followed by affinity capture of the RBP under study, using specific antibodies against the native protein itself or a tagged version of the RBP expressed in the cells. The isolated complex is then subjected to limited RNase digestion, radioactive end labeling of the bound RNAs, and size selection by denaturing gel electrophoresis. Finally, RNA recovered from the complexes is used for reverse transcription and next-generation sequencing. The RNase trimming step ensures that transcript regions bound by the bait protein will be preferentially sequenced, whereas the propensity of the PAR-cross-linking chemistry to induce thymidine (T) to cytidine (C) transitions during reverse transcription helps with precise binding site identification. Even if PAR-CLIP is currently very popular, combinations of UV-cross-linking, RNP immunoprecipitation, isolation of cross-linked RNA segments, and cDNA sequencing have been developed before (e.g., CLIP) (Ule 2003), and related methods using microarrays or high-throughput sequencing exist to profile RNAs associated with immunopurified RNA-binding proteins (RIP-Chip, RIP-Seq) (Tenenbaum et al. 2000).

Figure 2.

Methods to study mRNP composition. (A) Photoactivatable-ribonucleoside-enhanced cross-linking and immunoprecipitation (PAR-CLIP). Cells are cultured in media containing 4-thiouridine (4SU), leading to incorporation of the photoactivatable nucleoside into cellular RNA. Cross-linking with UV light of 365 nm leads to covalent attachment of RBPs to RNA targets that withstands partial RNAse digestion, immunoprecipitation, and purification by denaturing gel electrophoresis. Isolated RNA fragments are identified by next-generation sequencing, aided by a tendency of the cross-linked site to show thymidine (T) to cytidine (C) transitions. (B) GRNA chromatography using specific interaction between a 21-amino-acid peptide from the λ phage N anti-terminator protein and the boxB hairpin. A fusion of λN peptide with glutathione S-transferase (GST), and incorporation of the boxB hairpin into bait RNA converts glutathione Sepharose into an RNA affinity matrix (GRNA resin), which is incubated with cellular extracts. Proteins specifically bound to the matrix are eluted and identified by mass spectrometry. (C) Interactome capture. The procedure begins with RNP cross-linking in living cells by conventional UV 254-nm cross-linking or as in the PAR-CLIP approach. Following lysis, the complete cellular complement of (m)RNPs is purified by binding to an oligo(dT) resin and stringent washing under conditions that dissociate noncovalent RNA–protein interactions. Specifically bound proteins are released by RNase digestion and identified by mass spectrometry. (Diagrams are based on data from Czaplinski et al. 2005, Hafner et al. 2010, and Castello et al. 2012, respectively.)

Identification of Trans-Acting Factors by RNA Affinity Chromatography

The converse approach to CLIP experiments, using a given RNA under study as bait to purify and identify interacting proteins, is also commonly used. Here the RNA is typically engineered to contain a small sequence or structural tag that will facilitate specific capture of the mRNP complex. GRNA affinity chromatography is an example of such an approach (Fig. 2B) (Czaplinski et al. 2005; Duncan 2006). It uses the specific binding of an RNA hairpin to a short peptide from the N anti-terminator protein of the λ phage (termed BoxB hairpin and λN peptide, respectively, originally developed as a tool for tethered function assays) (De Gregorio et al. 1999). For GRNA chromatography, native mRNP complexes are assembled on boxB-tagged RNAs in vitro, which are then purified with the help of λ-GST fusion proteins bound to a solid support. Proteins that copurify with the tagged RNA are then identified by mass spectrometry. Besides the boxB/λ tether, several other combinations of RNA aptamers (e.g., the streptomycin or tobramycin tags) or tethering proteins (e.g., MS2 coat protein) have been used successfully (Beach and Keene 2008).

Although these approaches have been quite successful in vitro, only a few success stories have been reported from in vivo settings (Hogg and Collins 2007). Recently, the MS2 coat protein fused to a tag consisting of two His6 clusters separated by a cleavage site for the TEV protease and an in vivo biotinylation site has been used to capture IRES-binding proteins from 293 cells (Tsai et al. 2011). Furthermore, the development of “designer RNA-binding domains” that can be engineered to bind any desired target RNA sequence (Filipovska et al. 2011; Mackay et al. 2011) holds further promise for the purification of specific native mRNPs from living cells.

Discovery of RBPs by Interactome Capture

Given the high sensitivity of contemporary mass spectrometric approaches to determine thousands of proteins in complex mixtures, two groups recently achieved the first step in the quest for complete mRNA interactomes (see above), the identification of “all” mRNA-binding proteins of a mammalian cell (Castello et al. 2012). This work first used efficient in vivo cross-linking (conventional UV254 cross-linking, or PAR-CL) to preserve physiologically relevant RNA–protein interactions, followed by capture of the polyadenylated (m)RNAs with their cross-linked RBPs on an oligo(dT) matrix, stringent washing, subsequent release of bound proteins by RNase digestion, and finally identification by mass spectrometry (Fig. 2C). This approach extends very early work on hnRNP proteins in vivo (Dreyfuss et al. 1984) and recent studies that used hybridization of labeled mRNA preparations to protein arrays to identify cellular RNA-binding proteins in vitro (Scherrer et al. 2010; Tsvetanova et al. 2010). It also has the advantage of sensitively and selectively detecting the near-complete array of native protein–mRNA interactions as they occur in living cells. This interactome capture approach can now be used or adapted to study the mRNA interactomes of other cells and to investigate changes in interactome composition as a function of biological conditions, such as differences in cell growth or cell cycle phase, or forms of stress (hypoxia, oxidative stress, nutrient deprivation, etc.).

FROM COMPLEX RNPs TO DEFINED CIS/TRANS INTERACTIONS

Proteins rarely act alone to regulate translation. Rather, multi-subunit complexes assemble on (the UTRs of) the transcript, directed by interactions between the RBP components of the complex and regulatory cis-acting sequences or structures of the mRNA. These complexes may also contain regulatory RNAs (e.g., miRISC), using nucleic acid hybridization as a principle of site-specific binding. It has now become clear that components of regulatory complexes and regulatory factors (or the translational machinery) need to interact dynamically to achieve accurate translational control. In the following, we focus on three illustrative examples of coordinated translational regulation. Two of them involve the cooperation between multiple cis-elements and trans-acting factors to ensure tight and timely control of translation. In the third example, temporal control of translation is achieved by the stepwise assembly and activation of a regulatory complex that coordinates the expression of a posttranscriptional operon.

TIGHT TRANSLATIONAL REPRESSION, A MULTIFUNCTIONAL RBP AND COMBINATORIAL REGULATION: Msl2 mRNA

MSL2 is the limiting component of the Drosophila dosage compensation complex, a chromatin assembly that equalizes the expression of X-linked genes between males (XY) and females (XX) by promoting hypertranscription of the single male X chromosome (for review, see Gelbart and Kuroda 2009). Dosage compensation must be repressed in females for viability, and this is primarily achieved by preventing MSL2 expression. Two posttranscriptional control mechanisms cooperate to inhibit msl2, and both are exerted by the female-specific RBP Sex-lethal (SXL). In the nucleus, SXL binds to oligo-uridine stretches adjacent to the splice sites of a small facultative intron in the msl2 5′ UTR to inhibit its splicing; this splicing inhibition retains the SXL-binding sites in the mature mRNA. In the cytoplasm, SXL inhibits msl2 translation by binding to specific sites in both the retained intron and the 3′ UTR (for review, see Graindorge et al. 2011). Binding of SXL to both UTRs is necessary for tight translational repression, but partial inhibition can be achieved by each UTR alone, allowing the stepwise dissection of the mechanism (Bashaw and Baker 1997; Kelley et al. 1997; Gebauer et al. 1999). Extensive mutational and functional analyses have revealed that SXL regulates translation by a dual mechanism: SXL bound to the 3′ UTR inhibits the recruitment of the 43S ribosomal complex to the mRNA, whereas SXL bound to the 5′ UTR blocks the scanning of those complexes that have presumably escaped the 3′-UTR-mediated control (Fig. 3) (Gebauer et al. 2003; Beckmann et al. 2005). A uORF located just upstream of the SXL-binding site has been shown recently to be important for 5′-UTR-mediated repression (Medenbach et al. 2011); SXL promotes the recognition of the upstream initiator AUG by scanning 43S complexes, thus preventing them from reaching the main ORF. SXL does not simply act by steric hindrance, because PTB bound in the same position does not promote uAUG recognition. These data suggest that SXL establishes specific contacts with additional factors and/or the translational machinery for repression via the 5′ UTR; they also provide a first example of RBP regulation of a uORF.

Figure 3.

Mechanism of translational repression of msl2 mRNA. SXL binds to both the 5′ and 3′ UTRs of msl2 to achieve strong repression. SXL bound to the 3′ UTR recruits UNR to bind to the RNA in close proximity. In turn, UNR interacts with poly(A) tail-bound PABP to inhibit 43S ribosomal complex recruitment at a step downstream from closed-loop formation (1). SXL bound to the 5′ UTR inhibits ribosomal scanning by promoting recognition of an upstream AUG (uAUG), thus preventing 43S complexes from reaching the main msl2 ORF (2). Additional unidentified factors (X, Y) are likely involved. (Adapted in modified form from Graindorge et al. 2011.)

Translational repression via the 3′ UTR certainly requires additional factors. Binding of SXL alone is insufficient to repress the 43S recruitment step, and the highly conserved SXL homolog from Musca domestica cannot inhibit translation despite binding to the same sites with similar apparent affinities, suggesting that specific contacts are made between SXL, the msl2 3′ UTR, and other factors necessary for repression (Gebauer et al. 2003; Grskovic et al. 2003). One of the critical factors was identified as the protein Upstream of N-ras (UNR), a conserved regulator also known for its role in IRES-mediated translation and mRNA stability control in mammals (Fig. 3) (for review, see Mihailovich et al. 2010). UNR is required for translational repression of msl2 reporters in vitro, for repression of endogenous msl2 in cell culture, and for inhibition of dosage compensation in female flies (Abaza et al. 2006; Duncan 2006; Patalano et al. 2009). Binding of UNR to the 3′ UTR of msl2 depends on SXL, and therefore, even though UNR is present in males, it does not bind to msl2 and repress its translation because of the absence of SXL. Thus, SXL confers a sex-specific function to UNR.

How does the SXL–UNR complex function to repress translation? Although a poly(A) tail is not strictly required for regulation, translational repression via the 3′ UTR is stimulated by the poly(A) tail (Duncan et al. 2009). Interactions between UNR and poly(A) tail-bound PABP are thought to underlie this stimulation by mechanisms that are yet unknown. PABP binds to the poly(A) tail and contacts the cap-binding complex, yielding a closed-loop conformation of the mRNA that is thought to be optimal for efficient ribosome recruitment (Tarun and Sachs 1996; Wells et al. 1998). Intriguingly, closed-loop formation of the msl2 mRNA is not affected by UNR, indicating that the SXL–UNR complex inhibits ribosome recruitment by targeting a step in translation initiation that is downstream from eIF4F binding (Fig. 3) (Duncan et al. 2009). Close examination of the msl2 3′ UTR indicates the presence of sequences required for translational repression but dispensable for SXL–UNR binding, suggesting that the full 3′-UTR repressor complex contains additional factors (C Militti, E Szostak, and F Gebauer, unpubl.). Understanding the composition of this complex and the interactions of its components with the translational machinery may yield novel clues as to how the SXL/UNR-organized complex on the 3′ UTR of the msl2 mRNA controls the recruitment of ribosomes.

In summary, tight translational repression of msl2 mRNA is achieved through an elaborate combinatorial mechanism that involves targeting different steps of translation initiation from both ends of the mRNA, involving multiple RBPs and a uORF as an additional cis-regulatory element. How all of these elements act together to ensure efficient and coordinated repression warrants further investigation.

MORE COMPLEXITY FOR TEMPORAL AND SPATIAL CONTROL OF TRANSLATION: nanos mRNA

Nanos (Nos) is a posterior determinant required for the formation of abdominal segments during early Drosophila development. Synthesis of Nos exclusively at the posterior of the embryo is achieved by localization and translational activation of the nos transcript at this region, combined with translational repression elsewhere (Gavis and Lehmann 1994). nos mRNA is transcribed and actively translated in nurse cells, and subsequently transferred to the adjacent growing oocyte through ring canals. In the bulk cytoplasm of late oocytes and early embryos, nos mRNA is translationally repressed via sequences contained in a discrete region of the 3′ UTR proximal to the stop codon, the translational control element (TCE) (Dahanukar and Wharton 1996; Gavis et al. 1996; Smibert et al. 1996). The TCE consists of three stem–loops that are necessary for repression (Fig. 4). The AU-rich stem of one of these structures (stem IIIA, following the nomenclature of Crucs et al. 2000) is necessary for translational repression in oocytes and is thought to bind the hnRNP F/H protein Glorund (Crucs et al. 2000; Forrest et al. 2004; Kalifa et al. 2006). The other two stems carry loops consisting of CUGGC, which are recognized by the repressor Smaug and are therefore referred to as Smaug recognition elements (SREs) (Smibert et al. 1996, 1999; Dahanukar et al. 1999). Smaug is expressed exclusively in the early embryo and acts as a major regulator of maternal mRNA destabilization upon egg activation (Tadros et al. 2007). Point mutations in the SREs affect Smaug binding and lead to nos derepression in the embryo without effects on localization of the mRNA (Smibert et al. 1996). Smaug binds to the SRE via its SAM (sterile α motif) domain (Aviv et al. 2003; Green et al. 2003). A central guanine in the SRE appropriately oriented by the stem–loop structure is critical for SAM domain recognition (Johnson and Donaldson 2006; Oberstrass et al. 2006).

Figure 4.

Mechanism of translational repression of nanos mRNA. nanos mRNA switches from a translationally active state in nurse cells to a silenced state in late oocytes and early embryos. Repression in late oocytes is driven by Glorund (Glo) binding to stem IIIA of the TCE and seems to be effected primarily at the elongation step. In embryos, Smaug (Smg) is synthesized and takes over repression by binding to the SREs within the TCE; the relative contribution of each SRE is uncertain, although SRE1 seems to contribute more to Smaug binding than SRE2. Smg recruits the eIF4E-binding protein Cup and the CAF–CCR4–NOT complex to block translation initiation and promote deadenylation and degradation of nos mRNA. The piwi pathway has been recently reported to participate in deadenylation. Additional steps of translation could be affected both in late oocytes and early embryos (see text for details).

Smaug promotes deadenylation of nos mRNA by recruitment of the CAF–CCR4–NOT complex (Jeske et al. 2006; Zaessinger et al. 2006). Deadenylation is necessary for nos repression, but deadenylated nos reporters can still be strongly repressed in vitro in a manner that partially depends on the SREs (Jeske et al. 2006). These data suggest that Smaug-mediated repression has two components: one that is independent of the poly(A) tail, and one that depends on deadenylation. Indeed, translational repression is often coupled to deadenylation, which contributes to maintain the repressed state (for review, see Miller and Olivas 2011). Smaug interacts with Cup, a protein that binds eIF4E and blocks the recruitment of eIF4G to the cap complex (Fig. 4) (Nelson et al. 2004; for review, see Richter and Sonenberg 2005). Mutation of the eIF4E-binding motifs of Cup partially relieves repression of nos transgenes, and the binding of Cup and eIF4G to nos mRNPs is mutually exclusive, implying a function of Cup in mediating a translation initiation block by Smaug (Nelson et al. 2004; Jeske et al. 2011). Contradictory results have been reported concerning the requirement of the cap structure for Smaug-mediated repression, however (Andrews et al. 2011; Jeske et al. 2011). Furthermore, a significant fraction of nos mRNA was found in association with polysomes in early embryos even under conditions in which the mRNA is completely unlocalized and Nos protein is undetectable, suggesting that a postinitiation step is affected (Clark et al. 2000). Consistently, translation mediated by the Cricket paralysis virus (CrPV) IRES, which mediates translation initiation without requirement for any of the cellular translation initiation factors, can be inhibited by the SREs (Jeske et al. 2011). Therefore, repression by Smaug might involve initiation and postinitiation events. Interestingly, recent data suggest that Cup can elicit translational repression independent of its eIF4E-binding motifs, and that Cup also promotes deadenylation directly via association with the CAF–CCR4–NOT complex (Igreja and Izaurralde 2011). This raises the possibility that Cup mediates repressor functions of Smaug beyond translation initiation. It will be interesting to analyze to what extent Smaug function is preserved in Cup-depleted extracts.

Recently, late ovary extracts that recapitulate repression mediated by the IIIA stem (the Glorund-binding site) have been developed (Andrews et al. 2011). Repression seems to be cap independent in these extracts. In addition, nos mRNA is found associated with polysomes, and Glorund is present in polysomes in association with the repressed mRNA, suggesting that Glorund inhibits translation at a postinitiation step. On the other hand, repression in late oocytes is poly(A) dependent, which may reflect an effect of Glorund on initiation as well. A comparison of the polysomal association of nos mRNA in total ovary extracts, which are enriched for early-stage oocytes, with that in late ovary and embryo extracts indicates a gradual shift to lighter fractions, consistent with the temporal acquisition of distinct mechanisms of translational repression (Andrews et al. 2011).

Altogether, currently available data can be integrated into the following model explaining the temporal switch in nos mRNA expression, from activation in nurse cells to repression later in development. In late oocytes, Glorund imposes a block on elongation/termination that results in quick repression; in early embryos, Smaug consolidates nos inhibition at the initiation and possibly postinitiation steps and promotes deadenylation and degradation of the transcript by recruiting the CAF–CCR4–NOT complex (Fig. 4). Intriguingly, the piRNA pathway has recently been implicated in this mechanism. Rouget et al. (2010) found that CCR4-mediated deadenylation of nos depends on piRNAs complementary to a distal region of nos 3′ UTR and that Aubergine, an Argonaute protein, interacts with Smaug and CCR4. Further experiments are required to decipher how the piRNA pathway cooperates with the Smaug–CCR4 complex.

SIGNAL-DEPENDENT TEMPORAL CONTROL OF AN (ANTI-)INFLAMMATORY RNA OPERON: THE GAIT COMPLEX

During inflammation, the synthesis of ceruloplasmin (Cp) is transiently induced by interferon (IFN)-γ in myeloid cells and ceases at ∼24 h of IFN-γ treatment by the action of a translation repressor complex termed GAIT (IFN-γ-activated inhibitor of translation) (for review, see Mukhopadhyay et al. 2009). The GAIT complex recognizes a bipartite stem–loop structure in the 3′ UTR of Cp mRNA, the GAIT element (Fig. 5) (Sampath et al. 2003). The GAIT complex is composed of four proteins: ribosomal protein L13a, glutamyl-prolyl tRNA synthetase (EPRS), NS1-associated protein 1 (NSAP1), and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (Mazumder et al. 2003; Sampath et al. 2004). EPRS is the subunit that binds directly to the RNA, whereas phosphorylated L13a is responsible for interactions with the translational machinery (Sampath et al. 2004; Kapasi et al. 2007; Arif et al. 2009). Similar to 3′-UTR-bound SXL, the GAIT complex inhibits 43S recruitment without affecting closed-loop formation (Mazumder et al. 2001; Kapasi et al. 2007). Furthermore, because repression requires PABP and the poly(A) tail, it was proposed that mRNA circularization actually favors the correct positioning of GAIT close to the 5′ UTR (Mazumder et al. 2001). L13a binds eIF4G at its eIF3-binding site and blocks the eIF4G–eIF3 interaction, a step required for 43S complex recruitment (Kapasi et al. 2007; Arif et al. 2009).

Figure 5.

Translational repression by the GAIT complex. During inflammation, EPRS and L13a are phosphorylated and released from the multisynthetase complex (MSC) and the 60S ribosomal subunit, respectively. These proteins bind to NSAP1 and GAPDH to form the heterotetrameric GAIT complex, which binds to a split stem structure present in the 3′ UTR of target mRNAs. L13a then inhibits the recruitment of the 43S ribosomal complex by blocking the interaction between 43S-associated eIF3 and eIF4G.

Several lessons can be learned from this instructive example of translational control. First, temporal control of translation is achieved by the regulated, ordered assembly of the GAIT complex. Under control conditions, EPRS resides in the cytosolic tRNA multisynthetase complex (MSC) and acts as an enzyme catalyzing the addition of amino acids to tRNA. EPRS contains two catalytic domains (ERS and PRS) linked by a region that is not required for its enzymatic activity. This region contains three helix–turn–helix structures termed WHEP domains (named after the three tRNA synthetases that contain them) that serve RNA binding (Jia et al. 2008). Phosphorylation of EPRS by Cdk5 during the early phase of the IFN-γ response induces its release from the MSC (Sampath et al. 2004; Arif et al. 2009, 2011). Free, phosphorylated EPRS interacts with NSAP1 through one of the WHEP domains, resulting in a complex with no RNA-binding activity (Fig. 5). Later on, L13a—which resides in the large ribosomal subunit—is phosphorylated and released from the ribosome (Mazumder et al. 2003). Phosphorylated L13a and GAPDH then interact with the phosphorylated EPRS–NSAP1 dimer, resulting in a heterotetrameric complex that exposes the WHEP domains for interactions with the GAIT element (Jia et al. 2008).

The second lesson concerns the functional versatility of RBPs. As mentioned above, EPRS and GAPDH are polypeptides that can function as enzymes or contribute to translational control depending on cellular conditions. The founding example of an enzyme that also functions as an RBP is IRP1. In iron-replete cells, IRP1 bears an iron–sulfur cluster necessary for cytosolic aconitase activity. In iron-starved cells, the iron–sulfur cluster is lost, and a pocket is uncovered that binds to mRNAs encoding factors involved in iron metabolism, resulting in stabilization or translational repression of its target mRNAs (for review, see Pantopoulos 2004). Interconnectivity between cellular metabolism and RNA regulation might be an emerging theme in gene expression (Hentze and Preiss 2010).

The third lesson pertains to the observation that L13a phosphorylation leads to L13a-depleted ribosomes that are functionally normal. Practically the entire cellular complement of L13a is released upon phosphorylation, yielding no defects in general translation (Mazumder et al. 2003). This result suggests that the ribosome also serves as a repository of regulatory molecules for release from its surface when their functions are required. The concept of large macromolecular complexes as depots for regulatory proteins has been discussed elsewhere (Ray et al. 2007).

Finally, the fourth lesson relates to the existence of an RNA regulon for GAIT-mediated translational control during the inflammatory response. L13a phosphorylation is the culminating event of a kinase cascade in which IFN-γ activates DAPK, which, in turn, activates ZIPK (Fig. 5). Both the DAPK and ZIPK mRNAs contain GAIT elements in their 3′ UTRs and are repressed by the GAIT complex, establishing a negative-feedback loop that contributes to the temporal limitation of the inflammatory response (Mukhopadhyay et al. 2008). Furthermore, in silico searches and genome-wide polysome profiling have provided a list of candidate targets containing GAIT elements, many of which encode proteins involved in inflammation (Ray and Fox 2007; Vyas et al. 2009). These results imply that the GAIT complex appears to coordinate a posttranscriptional operon involved in the resolution of the inflammatory response.

CONCLUSIONS AND PERSPECTIVES

Starting with the mapping of cis-acting elements and the identification of trans-acting factors mostly by biochemical and genetic approaches during the past decades, the investigation of mRNA translation has now entered into a phase of transcriptome-wide, highly parallel analyses. These approaches have begun to help define RBP-binding sites across the transcriptome and paint a picture of dense mRNP assemblies involving a multitude of trans-acting factors. Several concepts have emerged along the way: (1) The binding sites of RBPs, and RBPs themselves, influence each other on the mRNAs, giving rise to combinatorial outcomes. (2) Essentially all RBPs have numerous target mRNAs, further driving combinatorial modes of translational control. (3) The resulting mRNPs are highly dynamic structures that undergo rearrangements in response to biological signaling. (4) DExH/D RNA helicases not only remodel RNA structure by unwinding RNA–RNA duplex structures, but can directly remodel RNPs in different biological settings (Jankowsky and Bowers 2006). Reductionist biochemical work on model systems revealed that the same RBPs can influence translation by more than one mechanism, often influenced by other RBPs in a combinatorial way. We expect that reductionist mechanistic investigations by biochemical approaches and transcriptome-wide, time-resolved in vivo analyses including ribosome profiling (Ingolia et al. 2011) will converge to yield unprecedented insights into translation and translational control, both at the level of individual mRNAs and whole transcriptomes.